Thanks. Had to watch this twice with starting and stopping to read the little comments... But after years of YUV mystifying me, I finally understand it. Great tutorial
If you're watching on TH-cam, the video is 8 bit YUV420, which is then converted to RGB for the phone screen. If you plug your phone into a TV, it depends. HDMI supports either scheme (but you should probably force it to use RGB)
The thing that is bugging me out, is that even if in theory, we are not suposed to notice 422 from 444, on my TV (an LG C9) the difference is pretty evident. 422 is definitely a bit blurrier, and 444 have razor sharp edges and retains more fine detail in textures. I'm starting to think that maybe I'm too sensitive to color?¿
For general text or graphics? I find it very noticeable as well - particularly coloured text on a background graphic. If there's a difference when watching movies or something however, that might come down to how the TV handles processing with different inputs. There are a couple other things with TVs: - Make sure your colour range is set to limited/full depending on how your TV works. Full uses 0-255 for representing each channel, but most TVs expect limited range 16-235. - If overscan is kicking in, it also makes everything very blurry since it scales (usually bilinear) the input
If you have a 4K 4:2:0 video from a DSLR, and re-encode at 1080p 4:4:4, would you have more informations than reencoding at 1080p 4:2:0 ? Well nothing would beat the original 4K obviously
Hm. In theory, yes, you should be able to then have 1080p luma and full 1080p chroma vs 1080p luma and 540p chroma. That being said, I have not looked into how FFMPEG/etc actually handle the chroma channels. I know that they do use different scaling algorithms for luma and chroma by default when transcoding, but if it'll work as you mentioned, I'm not sure.
Depends what you're trying to do and what your priorities are. As shown in the video, on PC displays, full 24 or 30 bit RGB would be the "best" video quality, but with terrible efficiency and be very wasteful. On the other hand, if your goal is compressing video down to manageable bitrates with something simple enough that it can be mostly decoded in hw, YUV420. In terms of workloads, videos and JPEG photographs are typically in YUV. Mostly for efficiency and due to legacy/backwards compatibility reasons. PNG (graphics) are stored in RGB, as they are often used for graphics or charts for web use, where you want something lossless. This is actually an ongoing discussion, as we need to decide whether to use YUV or RGB for new image formats such as JXL and AVIF. Photography tends to use YUV with lossy compression, but PC and web graphics tends to use RGB with lossless compression. AVIF for example, has a lossless mode, but, since it uses AV1 video I frames (simplifying a fair bit here), it converts an input image in RGB to YUV. This process however isn't lossless, and you can see color artifacts around edges of certain objects. TLDR - which is best depends on what you're doing, and what your goals are.
![ตีสิบเดย์ [FULL] | ตัวตึง! ปากกล้า! ท้ารบ! "ทนายไพศาล เรืองฤทธิ์"](http://i.ytimg.com/vi/pU3dXTsaBVc/mqdefault.jpg)
Thanks. Had to watch this twice with starting and stopping to read the little comments... But after years of YUV mystifying me, I finally understand it. Great tutorial
Thanks! Glad I could help demystify this a bit!
I feel like I can lecture people about this now, thx
Thank you !! GReat video
thanks, cool video!
4K colour space Rec. 2020 has a really wide RGB gamut too HENCE why it really is a huge upgrade
Please tell me whether I use RGB or yuv color format for watching content on phone? ❤
If you're watching on TH-cam, the video is 8 bit YUV420, which is then converted to RGB for the phone screen.
If you plug your phone into a TV, it depends. HDMI supports either scheme (but you should probably force it to use RGB)
The thing that is bugging me out, is that even if in theory, we are not suposed to notice 422 from 444, on my TV (an LG C9) the difference is pretty evident. 422 is definitely a bit blurrier, and 444 have razor sharp edges and retains more fine detail in textures.
I'm starting to think that maybe I'm too sensitive to color?¿
For general text or graphics? I find it very noticeable as well - particularly coloured text on a background graphic. If there's a difference when watching movies or something however, that might come down to how the TV handles processing with different inputs.
There are a couple other things with TVs:
- Make sure your colour range is set to limited/full depending on how your TV works. Full uses 0-255 for representing each channel, but most TVs expect limited range 16-235.
- If overscan is kicking in, it also makes everything very blurry since it scales (usually bilinear) the input
If you have a 4K 4:2:0 video from a DSLR, and re-encode at 1080p 4:4:4, would you have more informations than reencoding at 1080p 4:2:0 ?
Well nothing would beat the original 4K obviously
Hm. In theory, yes, you should be able to then have 1080p luma and full 1080p chroma vs 1080p luma and 540p chroma.
That being said, I have not looked into how FFMPEG/etc actually handle the chroma channels. I know that they do use different scaling algorithms for luma and chroma by default when transcoding, but if it'll work as you mentioned, I'm not sure.
YUV vs RGB wich best?
Depends what you're trying to do and what your priorities are. As shown in the video, on PC displays, full 24 or 30 bit RGB would be the "best" video quality, but with terrible efficiency and be very wasteful. On the other hand, if your goal is compressing video down to manageable bitrates with something simple enough that it can be mostly decoded in hw, YUV420.
In terms of workloads, videos and JPEG photographs are typically in YUV. Mostly for efficiency and due to legacy/backwards compatibility reasons. PNG (graphics) are stored in RGB, as they are often used for graphics or charts for web use, where you want something lossless.
This is actually an ongoing discussion, as we need to decide whether to use YUV or RGB for new image formats such as JXL and AVIF. Photography tends to use YUV with lossy compression, but PC and web graphics tends to use RGB with lossless compression. AVIF for example, has a lossless mode, but, since it uses AV1 video I frames (simplifying a fair bit here), it converts an input image in RGB to YUV. This process however isn't lossless, and you can see color artifacts around edges of certain objects.
TLDR - which is best depends on what you're doing, and what your goals are.